模拟电路自动设计的多目标自适应进化方法  

Evolutionary Design of Analog Circuits Using a Multi-objective Adaptive Genetic Algorithm

作　　者：赵曙光[1] 刘贵喜[1] 杨万海[1] 

机构地区：[1]西安电子科技大学电子工程学院,西安710071

出　　处：《仪器仪表学报》2002年第z3期90-93,共4页Chinese Journal of Scientific Instrument

基　　金：陕西省自然科学基金资助项目 ( 99X0 8)

摘　　要：滤波器等模拟电路在仪器仪表中起着不可替代的作用 ,但手工设计步骤繁琐且精度较低。进化设计可实现电路自动设计和优化。在此所提出的改进遗传算法 ,利用 :支持元件种类变化和取值规格化的编码——解码方案 ,基于 Pspice仿真的多目标评估方法 ,基于平均适应度的遗传参数自适应策略等 ,可自动调整电路结构和优化元件取值 ,得到具备预期特性的电路 ,并且显著提高进化的速度、规模和实用性。有源滤波器等实验结果证明了该方法的有效性。Analog circuits are of great importance in electronic system,especially for instruments and apparatus.Owing to difficulties of analog circuit design and the shortage of qualified analog design engineers,evolutionary computation based techniques are expected to design circuits automatically. In belief that some rules and experiences of the conventional method will be helpful to change this status,we propose a novel genetic algorithm adapted for automated circuit design. It is characterized by three features:a encoding-decoding scheme allowing for selection of type and use of preferred value (e.g.a 24 series standard value) of components, multi-objective fitness evaluation based on PSPICE simulation,and average fitness based genetic parameters adaptation. The experiment results of active filters showed that our approach is able to select a suitable circuit structure and generate a feasible combination of components both in types and values,result in a response close to the design objects.

关 键 词：遗传算法 电路进化设计 可进化硬件 

分 类 号：TP2[自动化与计算机技术—检测技术与自动化装置]